<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> ArgusBot is a Python supervisor plugin for Codex CLI and Claude Code CLI, designed to overcome the common issue of AI agents stopping prematurely. It provides a 24/7 supervisory agent that ensures tasks are executed, reviewed, and planned until completion, offering a robust solution for automated workflows and research.

How It Works

The core mechanism is a persistent loop: a main agent executes tasks via a chosen backend (Codex, Claude, or GitHub Copilot CLI), a reviewer sub-agent assesses task completion (done/continue/blocked), and a planner sub-agent maintains state and proposes next steps. This cycle continues until the reviewer signals completion and all acceptance criteria are met, preventing premature agent termination.

Quick Start & Requirements

• Installation: Clone the repo and install: git clone <repo_url>; cd ArgusBot; python -m pip install -e .
• Prerequisites: Python, authenticated Codex CLI or Claude Code CLI. Optional: copilot-proxy, GitHub Copilot CLI, Telegram/Feishu credentials.
• Setup: Run argusbot init in your target project directory.
• Resource Notes: High reasoning levels incur significant token costs; medium reasoning offers a good quality/cost tradeoff.
• Documentation: Links to "Quick Start", "Recent Updates", and "Contributing" are referenced.

Highlighted Details

• Persistent Agent Loop: Ensures task completion via continuous execution, review, and planning.
• Dual Control Channels: 24/7 remote operation via Telegram (with voice/audio transcription) or Feishu.
• Copilot Proxy Integration: Routes Codex backend calls through copilot-proxy for GitHub Copilot quota usage.
• Native GitHub Copilot CLI Backend: Direct support for GitHub Copilot CLI.
• Advanced Planning: Features a planner with live plan snapshots and follow-up objective proposals.
• Visibility: Terminal streaming, local dashboard, Telegram/Feishu push notifications.
• Resumability: Run archives (JSONL) enable continuity and auditing.

Maintenance & Community

Contributions are welcome via CONTRIBUTING.md. A user community for research workflows is available.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

• Security Risks: Daemon runs default to --yolo (high local privileges); use only in trusted environments. Bot tokens require careful handling.
• Cost Management: Extended loops can lead to substantial token consumption; monitoring is advised.
• Debugging: Remote interfaces may obscure details; local terminal output is recommended for debugging.